How to Create an Interaction Plot in SPSS

What Is an Interaction Plot?

An interaction plot (also called a profile plot in SPSS) is a line graph that shows how the effect of one factor on your outcome changes across the levels of another factor — the clearest way to see whether two variables interact.

Table of Contents

If you have not run the analysis yet, work through our two-way ANOVA tutorial first — the plot in this guide is built from that analysis.

Imagine you are testing whether a new teaching method improves exam scores, and you also want to know whether it works differently for morning and evening classes. A table of means gives you the numbers, but a plot shows the story at a glance: if the lines for morning and evening classes run roughly parallel, there is probably no interaction. If they cross, fan out, or converge, the method’s effect depends on the time of day — that is an interaction.

When Should You Plot an Interaction?

Create an interaction plot whenever your analysis includes two or more factors and you want to understand how they combine. Typical situations include:

Two-way ANOVA: two categorical predictors (for example, teaching method × class time) and one continuous outcome.
Repeated-measures ANOVA: to see how scores change over time in different groups.
ANCOVA: to display adjusted means or check visually whether group differences are consistent across a covariate.
Presenting results: a plot communicates an interaction far faster than a table of cell means, and reviewers expect one when an interaction is significant.

A plot is useful even when the interaction is not significant: parallel lines are themselves a finding.

What You Need Before You Start

You need three variables at minimum:

One continuous outcome variable — for example, exam score.
Two categorical factors — for example, teaching method (traditional vs interactive) and class time (morning vs evening), each coded with numeric values and value labels.
No empty cells: every combination of the two factors should have at least a few cases, or SPSS will warn you and the plot will have gaps.

Before plotting, run the two-way ANOVA itself so you know whether the interaction is statistically significant — the plot shows you the pattern, but the ANOVA table gives you the p-value and effect size that belong in your write-up.

Method 1: Profile Plot Inside a Two-Way ANOVA (Recommended)

This is the fastest route, because the plot is produced alongside the ANOVA output you need anyway.

Step 1: Open the Univariate dialog

Click Analyze > General Linear Model > Univariate.

Step 2: Assign your variables

Move your outcome variable (for example, Exam Score) into the Dependent Variable box.
Move your two categorical variables (for example, Teaching Method and Class Time) into the Fixed Factor(s) box.

Step 3: Request the profile plot

Click the Plots button. In the Profile Plots dialog:

Move one factor (for example, Class Time) into the Horizontal Axis box.
Move the other factor (for example, Teaching Method) into the Separate Lines box.
Click Add. You should now see an entry such as Class Time*Teaching Method in the Plots list.
Click Continue.

Which factor goes on the horizontal axis is a presentational choice, not a statistical one — the interaction is the same either way. As a rule of thumb, put the factor with more levels, or the one that represents time or order, on the horizontal axis.

Step 4: Request the statistics you need

Click Options and tick Descriptive statistics and Estimates of effect size. These give you the cell means and partial eta squared that belong in your results section. Click Continue, then OK to run the analysis.

Step 5: Find the plot in the output

Scroll to the bottom of the Output Viewer. After the Tests of Between-Subjects Effects table you will see a section titled Profile Plots containing your interaction plot: the outcome on the vertical axis, one factor along the horizontal axis, and one line per level of the second factor.

How to Interpret an Interaction Plot

Read the plot by looking at the shape of the lines, not at any single point:

Parallel (or near-parallel) lines: the effect of one factor is roughly the same at every level of the other factor, so there is probably no interaction.
Lines that fan out or converge: the effect of one factor gets stronger or weaker depending on the other factor — an ordinal interaction when the lines do not cross.
Lines that cross: the effect of one factor reverses direction across levels of the other factor. This is a disordinal (crossover) interaction, and it matters most, because the main effects can be misleading when lines cross — for example, the interactive method might beat the traditional method in the morning but lose to it in the evening.

Always check the plot against the ANOVA table. A dramatic-looking crossover based on small cell sizes may not be significant, and a modest fan shape may be highly significant with a large sample. Report both: the pattern you can see and the test statistic that backs it up.

Also check the cell means in the Descriptive Statistics table. If one cell has very few cases, its point on the plot is unreliable — a single outlier can bend a line.

Method 2: Interaction Plot with the Chart Builder

Use the Chart Builder when you want a publication-ready figure — with error bars, custom colours, or a cleaner layout than the default profile plot. The Chart Builder does not run the ANOVA for you, so run the analysis first and use this route for the figure alone.

Step 1: Open the Chart Builder

Click Graphs > Chart Builder. If SPSS asks you to define measurement levels, click OK.

Step 2: Choose a line chart

In the Gallery tab, select Line, then drag the Multiple Line icon (the one showing several lines) onto the chart preview canvas.

Step 3: Assign the variables

Drag your outcome variable onto the Y-Axis drop zone. By default SPSS plots the mean — check the statistic shown and change it to Mean if needed.
Drag your first factor onto the X-Axis drop zone.
Drag your second factor onto the Set Color (grouping) drop zone at the top right of the canvas.

Step 4: Add error bars (optional but recommended)

Click the Element Properties tab, tick Display error bars, and choose 95% confidence intervals. Click Apply, then OK.

Step 5: Polish the figure

Double-click the chart in the Output Viewer to open the Chart Editor. From here you can change line colours and markers, edit the axis titles, add a legend title, and remove the grey background. When the figure looks right, right-click it and choose Export to save it as a PNG or EMF file for your thesis or manuscript.

SPSS Syntax for an Interaction Plot

If you prefer syntax — or you want a reproducible record of exactly what you ran — the profile plot from Method 1 is produced with the /PLOT subcommand in UNIANOVA. Replace the variable names with your own:

UNIANOVA ExamScore BY Method Time
  /METHOD=SSTYPE(3)
  /PLOT=PROFILE(Time*Method)
  /PRINT=DESCRIPTIVE ETASQ
  /DESIGN=Method Time Method*Time.

What each line does:

UNIANOVA ExamScore BY Method Time runs a two-way ANOVA with ExamScore as the outcome and Method and Time as factors.
/PLOT=PROFILE(Time*Method) requests the interaction plot, with Time on the horizontal axis and a separate line for each level of Method.
/PRINT=DESCRIPTIVE ETASQ adds the cell means and partial eta squared.
/DESIGN=Method Time Method*Time specifies the two main effects and their interaction.

To create a line chart of cell means through syntax instead, use GGRAPH with the Chart Builder’s GPL code, or use GRAPH /LINE(MULTIPLE) with MEAN(ExamScore) BY Time BY Method for a quick legacy plot.

Common Mistakes to Avoid

Plotting raw data instead of means: an interaction plot shows cell means, not individual scores. If your lines look jagged, check that the plot is summarising means.
Swapping the axes without thinking: putting the wrong factor on the horizontal axis does not change the statistics, but it can make the pattern harder to read. Try both arrangements and keep the clearer one.
Ignoring empty cells: if a factor combination has no cases, SPSS leaves a gap in the line. Check your cell counts in the Descriptive Statistics table first.
Reporting the plot without the test: a plot is a visual aid, not evidence on its own. Always pair it with the F statistic, degrees of freedom, p-value, and partial eta squared from the ANOVA table.
Over-interpreting non-significant patterns: lines that look slightly non-parallel in a small sample usually mean nothing. Let the interaction term in the ANOVA table decide whether the pattern is worth discussing.

Reporting in APA Style (7th ed.)

When an interaction is significant, report the ANOVA result, describe the pattern, and refer the reader to the figure. For example:

“A 2 × 2 between-subjects ANOVA revealed a significant interaction between teaching method and class time, F(1, 76) = 8.42, p = .005, partial η² = .10. As shown in Figure 1, the interactive method produced higher exam scores than the traditional method in morning classes, whereas the two methods did not differ in evening classes.”

Your figure needs a number and a clear title in APA style, placed above the figure, for example: Figure 1. Mean Exam Score by Teaching Method and Class Time. A brief figure note below the plot should state what the error bars represent, for example: Note. Error bars represent 95% confidence intervals.

If the interaction is not significant, say so plainly and report the main effects instead: “The interaction between teaching method and class time was not significant, F(1, 76) = 0.84, p = .362, partial η² = .01, indicating that the effect of teaching method did not differ between morning and evening classes (see Figure 1).”

Do not report exact p-values as “.000” — write p < .001. Report effect sizes (partial eta squared) for every effect you discuss, and give the cell means and standard deviations in the text or in a table so the reader can see the numbers behind the lines.

Key Takeaways

An interaction plot (SPSS profile plot) shows cell means as lines, making it easy to see whether one factor’s effect depends on another.
Parallel lines suggest no interaction; fanning, converging, or crossing lines suggest an interaction — confirm it with the ANOVA table, not by eye alone.
The quickest route is Analyze > General Linear Model > Univariate, then Plots, with one factor on the horizontal axis and the other as separate lines.
For a publication-ready figure, rebuild the plot in the Chart Builder and add 95% confidence-interval error bars.
Always report the interaction test (F, df, p, partial η²) alongside the figure, and describe the pattern in words — never let the plot stand alone.

Frequently Asked Questions

Can I create an interaction plot for a repeated-measures ANOVA?

Yes. In Analyze > General Linear Model > Repeated Measures, click Plots after defining your within-subjects factor, put the within-subjects factor (for example, time point) on the horizontal axis and the between-subjects factor (for example, group) in Separate Lines, then click Add. Non-parallel lines mean the groups change differently over time.

Should the interaction plot show raw means or estimated marginal means?

For a balanced design with equal cell sizes, raw cell means and estimated marginal means are identical. With unequal cell sizes, estimated marginal means are usually better, because they adjust for the imbalance. The default profile plot shows estimated marginal means; the Chart Builder shows raw means unless you change the statistic.

How do I know if the interaction in my plot is statistically significant?

You cannot tell from the plot alone. Look at the interaction row (for example, Method * Time) in the Tests of Between-Subjects Effects table. If its p-value is below your alpha level (usually .05), the interaction is significant. Report the F statistic, degrees of freedom, p-value, and partial eta squared with the figure.

Need this done on your data? Get a free quote → /services/